CN110223255B - Low-dose CT image denoising and recursion method based on residual error coding and decoding network - Google Patents
Low-dose CT image denoising and recursion method based on residual error coding and decoding network Download PDFInfo
- Publication number
- CN110223255B CN110223255B CN201910499262.XA CN201910499262A CN110223255B CN 110223255 B CN110223255 B CN 110223255B CN 201910499262 A CN201910499262 A CN 201910499262A CN 110223255 B CN110223255 B CN 110223255B
- Authority
- CN
- China
- Prior art keywords
- image
- network
- dose
- recursion
- low
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
- 238000000034 method Methods 0.000 title claims abstract description 21
- 238000013507 mapping Methods 0.000 claims abstract description 15
- 238000012549 training Methods 0.000 claims abstract description 14
- 230000008569 process Effects 0.000 claims abstract description 7
- ORILYTVJVMAKLC-UHFFFAOYSA-N Adamantane Natural products C1C(C2)CC3CC1CC2C3 ORILYTVJVMAKLC-UHFFFAOYSA-N 0.000 claims description 3
- 230000004913 activation Effects 0.000 claims description 2
- 238000005457 optimization Methods 0.000 claims description 2
- 238000011176 pooling Methods 0.000 claims description 2
- 238000012545 processing Methods 0.000 abstract description 2
- 230000000717 retained effect Effects 0.000 abstract 1
- 238000002591 computed tomography Methods 0.000 description 46
- 238000013528 artificial neural network Methods 0.000 description 4
- 230000006870 function Effects 0.000 description 4
- 238000013527 convolutional neural network Methods 0.000 description 3
- 238000002059 diagnostic imaging Methods 0.000 description 3
- 230000000694 effects Effects 0.000 description 3
- 230000009467 reduction Effects 0.000 description 3
- 238000003745 diagnosis Methods 0.000 description 2
- 238000010586 diagram Methods 0.000 description 2
- 238000002474 experimental method Methods 0.000 description 2
- 238000012805 post-processing Methods 0.000 description 2
- 238000012360 testing method Methods 0.000 description 2
- 230000018199 S phase Effects 0.000 description 1
- 230000003187 abdominal effect Effects 0.000 description 1
- 230000003044 adaptive effect Effects 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000013135 deep learning Methods 0.000 description 1
- 238000001514 detection method Methods 0.000 description 1
- 230000008034 disappearance Effects 0.000 description 1
- 238000004880 explosion Methods 0.000 description 1
- 238000001914 filtration Methods 0.000 description 1
- 238000003709 image segmentation Methods 0.000 description 1
- 230000005865 ionizing radiation Effects 0.000 description 1
- 230000009191 jumping Effects 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
- 230000001575 pathological effect Effects 0.000 description 1
- 230000005855 radiation Effects 0.000 description 1
Images
Classifications
-
- G06T5/70—
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T11/00—2D [Two Dimensional] image generation
- G06T11/003—Reconstruction from projections, e.g. tomography
- G06T11/008—Specific post-processing after tomographic reconstruction, e.g. voxelisation, metal artifact correction
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/10—Image acquisition modality
- G06T2207/10072—Tomographic images
- G06T2207/10081—Computed x-ray tomography [CT]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20081—Training; Learning
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20084—Artificial neural networks [ANN]
Abstract
The invention belongs to the technical field of image processing, and particularly relates to a low-dose CT image denoising recursive algorithm based on a residual error coding and decoding network; the specific technical scheme is as follows: a shallow residual coding and decoding recursion network for denoising a low-dose CT image is disclosed, wherein the recursion shallow residual coding and decoding network reduces the complexity of the network by reducing the number of layers and convolution kernels in the residual coding and decoding network, the performance of the network is improved by utilizing a recursion process, the algorithm learns end-to-end mapping through network training to obtain a high-quality image, and an original low-dose CT image is cascaded to the next input during each recursion, so that the problem of distortion of the image after multiple recursions can be effectively avoided, image characteristics can be better extracted, and detail information of the image is retained.
Description
Technical Field
The invention belongs to the technical field of image processing, and particularly relates to a low-dose CT image denoising and recursion algorithm based on a residual coding and decoding network.
Background
X-ray Computed Tomography (CT) provides major anatomical and pathological information of the human body for medical diagnosis and treatment. However, ionizing radiation generated during CT scanning can be harmful to the patient's body and may even cause canceration, so low dose CT applications are viable. A simple and easy to operate method of reducing radiation dose is to reduce the tube current, but this method can result in a reduction in the signal-to-noise ratio of the projection data, so that CT images reconstructed using the Filtered Back Projection (FBP) algorithm contain significant bar artifacts and noise, which affect the diagnosis of the physician. There has been considerable interest in how to reconstruct high quality CT images from raw noisy projection data.
The existing treatment methods mainly comprise three main types: a projection domain filtering algorithm, an iterative reconstruction algorithm and a post-processing method. Both of the first two algorithms process raw projection data, but researchers are often hindered due to the difficulty in acquiring raw data. The post-processing method is to directly process the reconstructed CT image, and the obtained algorithm can be easily implanted into the CT system and is of great interest.
Deep learning is used for image segmentation, super-resolution reconstruction, target detection and recognition due to its capability of extracting features, and is also used in the aspect of denoising low-dose CT images in recent years. A wavelet domain depth Convolution Neural Network (CNN) is proposed in an article "A deep conditional neural network using directional wavelets for low-dose X-ray CT reconstruction" (low-dose CT reconstruction based on a directional wavelet depth convolution neural network) published by Kang Eunhee in 2017 in Medical Physics (Medical Physics) journal. In 2017, chen hu proposed a shallow CNN of an image domain in a biological optical Express (Low-dose CT based on convolutional neural network) which shows that CNN can directly learn the end-to-end nonlinear mapping from a Low-dose image to a standard-dose image. The designed layer number is shallow, because the traditional CNN is easy to have gradient disappearance or explosion when the network deepens, so that the network performance is degraded, and in order to solve the problem, a residual error network is constructed.
The article "Wavelet residual network for Low-Dose CT via discrete conditional frames" (Wavelet residual network depth convolution framework for Low Dose CT) published by Kang Eunhee in the IEEE Transactions on Medical Imaging journal of 2017 applies 24 layers of residual networks for denoising over the Wavelet domain. An article "Low-Dose CT with a residual encoder-decoder connected neural network (RED-CNN)" (residual codec convolutional neural network of Low-Dose CT) published by Chen hu in the journal of IEEE Transactions on Medical Imaging (journal of institute of electrical and electronics engineers) in 2017 designs a residual codec network (RED-CNN) and applies to denoising of Low-Dose CT images. The network includes 10 layers, including 5 convolutional layers and 5 deconvolution layers. Each layer in the first 9 layers is provided with 96 convolution kernels, and the last layer is subjected to primary convolution by one convolution kernel to obtain a denoised CT image. Wolterink Jelmer2017 uses a Generative adaptive Networks for Noise Reduction in Low-Dose CT (Low Dose CT Noise Reduction Generative countermeasure network) for Low Dose CT image denoising, published in the journal IEEE Transactions on Medical Imaging, which includes both a Generative network with a convolutional neural network as a generator and an countermeasure network that optimizes the generator.
Disclosure of Invention
In order to solve the technical problem of complex network structure in the prior art, the invention designs a shallow recursive network, which recurs a constructed shallow residual coding and decoding network, constructs a new network by utilizing the same network structure recursion, reduces the complexity of the network by reducing the number of layers and the number of convolution kernels in the residual coding and decoding network, and achieves the purpose of obtaining high-quality images by utilizing a recursion process.
In order to realize the purpose, the technical scheme adopted by the invention is as follows: a low-dose CT image denoising and recursion method based on a residual error coding and decoding network comprises the following specific operation steps:
step one, arranging shallow layer residual error coding and decoding network
The network structure comprises 8 layers, wherein the 8 layers are composed of 4 convolution layers and 4 deconvolution layers which are symmetrically arranged, a ReLU activation function is arranged behind each convolution layer, and the network structure removes the pooling layer, so that the loss of structural details can be avoided. The first 4 convolutional layers constitute a stack encoder, which aims to remove noise and artifacts in the image; the last 4 deconvolution layers constitute a stack decoder, whose purpose is to restore the structural details of the image. Although the deconvolution layer can restore the details of the image, in order to better retain the detail information of the image, the residual error network is applied to the network, and the purpose of optimizing the mapping relationship of the network is achieved. Each convolution layer connects the extracted features to the symmetrical deconvolution layer in a jumping mode, so that the effect of keeping image details is achieved, and network training is facilitated. The first seven layers are all provided with 64 convolution kernels, the last layer is provided with one convolution kernel, and the denoised image is obtained through convolution.
Step two, recursive shallow residual error coding and decoding network
Through a shallow coding and decoding network, the image denoising effect is not ideal, so the invention introduces a recursive network. In each recursion, the original low-dose CT image and the denoised image after the previous recursion are cascaded to be used as the input of the next recursion, so that the problem of distortion of the image after multiple recursions can be avoided, and the detail characteristics of the original input image can be better extracted. The recursive process of the network can be expressed as:
wherein S is recursion times, X is network input, RED-Net is arranged residual coding and decoding network, and O s For the denoised CT image obtained in the s-th recursion, f in Is the output of the s recursion O s X-Cascade operation with original Low-dose CT image, I s+1 Is the input for the s +1 th recursion.
Step three, designing a loss function
The residual error coding and decoding recursive network is end-to-end mapping from a low-dose CT image to a standard metering image, and the mapping F is learned through the network;
given a training data set D = { (x) 1 ,y 1 ),(x 2 ,y 2 ),Λ,(x N ,y N ) In which { x } i I =1,2, Λ, N are patches extracted from low-dose CT images, { y } i I =1,2, Λ, N is the patch of images extracted from a standard dose CT image, N is the total number of training samples;
the parameters in the mapping F can be obtained by minimizing the following objective function
And step four, selecting an optimization algorithm, and optimizing by adopting an Adam algorithm.
And step five, extracting an image block set from a data set containing the standard dose CT image and the low dose CT image.
And step six, training a network through a data set to obtain a mapping relation F from a low-dose image to a standard-dose image.
And inputting the low-dose CT image into a designed residual error coding and decoding recursive network to obtain a final de-noised image.
Compared with the prior art, the invention has the following beneficial effects: the invention recurses the constructed shallow layer residual error coding and decoding network and utilizes the same network structure to recursion construct a new network. The complexity of the network is reduced by reducing the number of layers and the number of convolution kernels in the residual error coding and decoding network, and the purpose of obtaining high-quality images is achieved by utilizing a recursion process. In each recursion, the original low-dose CT image and the denoised image after the previous recursion are cascaded to be used as the input of the next recursion, so that the problem of distortion of the image after multiple recursions can be avoided, and the detail characteristics of the original input image can be better extracted. The invention not only reduces the complexity of the network, but also improves the network performance, so that the image details are well kept in the denoised image, and the structure is clearer.
Drawings
Fig. 1 is a structural relationship diagram of a shallow residual coding and decoding network.
Fig. 2 is a diagram of a RRED-Net network architecture with S-phase recursion.
Fig. 3 is a low dose CT image.
Fig. 4 is a standard dose CT image.
FIG. 5 is a resulting image after denoising by the method of the present invention.
FIG. 6 is a result image after RED-CNN denoising.
Fig. 7 is an enlarged view of the area within the box in fig. 3.
Fig. 8 is an enlarged view of the area within the box in fig. 4.
Fig. 9 is an enlarged view of the area within the box in fig. 5.
Fig. 10 is an enlarged view of the area within the box in fig. 6.
Detailed Description
In order to make the technical problems, technical solutions and advantageous effects to be solved by the present invention more clearly apparent, the present invention is further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention.
As shown in fig. 1 and fig. 2, a residual codec network-based low-dose CT image denoising and recursive method uses a clinical data set of "NIH-AAPM-Mayo Clinic low-dose CT big challenge in 2016 granted by Mayo clinical" for training and testing the proposed network, the data set includes standard dose and low-dose abdominal CT images of 10 anonymous patients, the size of the images is 512 × 512 pixels, and 3mm CT images are used for network training. In the experiment, the data used for training and testing is a fixed-size image block set extracted from a low-dose CT image and a standard-dose CT image, so that on one hand, the local features of the images can be effectively extracted, on the other hand, the data set can be increased, and overfitting is avoided.
The parameters set in the experiment were as follows: the size of the image block is composed of 54 × 54 pixels, and the learning rate α =10 -4 The number of cycles S =3, the number of layers of the codec network is 8, the number of convolution kernels in the last layer is 1, and the number of convolution kernels in the other layers is 64. The convolution kernel size of all layers consists of 5*5 pixels, the step size of convolution and deconvolution is set to 1 without zero padding, and the convolution and deconvolution kernels are initialized with a random gaussian distribution with a mean of 0 and a standard deviation of 0.01.
The proposed network is an end-to-end mapping from low dose CT images to standard dose images, with the mapping F learned over the network. Given a training data set D = { (x) 1 ,y 1 ),(x 2 ,y 2 ),Λ,(x N ,y N ) In which { x } i I =1,2, Λ, N are patches extracted from low-dose CT images, { y } i I =1,2, Λ, N is from a standard agentMeasuring image blocks extracted from the CT image, wherein N is the total number of training samples; the parameters in the mapping F can be obtained by minimizing the following objective function
And (3) obtaining a mapping relation F from a low-dose image to a standard-dose image through network training by adopting an Adam algorithm, and then inputting the low-dose CT image into a designed residual coding and decoding recursive network to obtain a final de-noised image.
In order to verify the performance of the network proposed by the present invention, we mark the network constructed by the present invention as RRED-Net and compare it with RED-CNN network, as shown in fig. 3, fig. 4, fig. 5, fig. 6, fig. 7, fig. 8, fig. 9 and fig. 10, it can be seen that the method RRED-Net and RED-CNN network of the present invention can effectively remove the artifacts in the image.
The invention makes comparison from the quantitative aspect, as shown in table 1, it can be seen that both methods obtain higher peak signal-to-noise ratio (PSNR) and Structural Similarity (SSIM), and the two indexes of the method provided by the invention are slightly higher than RED-CNN.
| PSNR | SSIM | |
| RED-CNN | 45.0123 | 0.9819 |
| RRED-Net | 45.0530 | 0.9821 |
Table 1: comparison of PSNR with SSIM
The above description is only for the purpose of illustrating the preferred embodiments of the present invention and is not to be construed as limiting the invention, and any modifications, equivalents and improvements made within the spirit and principles of the present invention are intended to be included therein.
Claims (1)
1. A low-dose CT image denoising and recursion method based on a residual error coding and decoding network is characterized by comprising the following specific operation steps:
step one, arranging shallow layer residual error coding and decoding network
The network structure comprises 8 layers, each layer consists of 4 convolution layers and 4 deconvolution layers which are symmetrically arranged, a ReLU activation function is arranged behind each convolution layer, the pooling layer is removed from the network structure, the 4 convolution layers form a stack encoder, the 4 deconvolution layers form a stack decoder, each convolution layer connects the extracted feature jump to the symmetrical deconvolution layers, 64 convolution kernels are arranged in the first seven layers, a convolution kernel is arranged in the last layer, and a denoised image is obtained by convolution in the last layer;
step two, recursive shallow residual error coding and decoding network
In each recursion, the original low-dose CT image and the denoised image after the last recursion are cascaded as the input of the next recursion, and the recursion process of the network is represented as follows:
wherein S is recursion times, X is network input, RED-Net is arranged residual coding and decoding network, and O s For the denoised CT image obtained in the s-th recursion, f in Is the output of the s recursion O s X-Cascade operation with original Low-dose CT image, I s+1 For the (s + 1) th recursionI.e. the cascaded images;
step three, designing a loss function
The residual error coding and decoding recursion network is end-to-end mapping from a low-dose CT image to a standard-dose image, and the mapping F is learned through the network;
given a training data set D = { (x) 1 ,y 1 ),(x 2 ,y 2 ),Λ,(x N ,y N ) In which { x } i I =1,2, Λ, N are patches extracted from low-dose CT images, { y } i I =1,2, Λ, N is the patch of images extracted from a standard dose CT image, N is the total number of training samples;
the parameters in the mapping F are obtained by minimizing the following objective function
Step four, selecting an optimization algorithm
Optimizing by adopting an Adam algorithm;
step five, extracting an image block set from a data set containing a standard dose CT image and a low dose CT image;
step six, obtaining a mapping relation F from a low dose image to a standard dose image through a data set training network;
and inputting the low-dose CT image into a designed recursive residual error coding and decoding network to obtain a final de-noised image.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910499262.XA CN110223255B (en) | 2019-06-11 | 2019-06-11 | Low-dose CT image denoising and recursion method based on residual error coding and decoding network |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910499262.XA CN110223255B (en) | 2019-06-11 | 2019-06-11 | Low-dose CT image denoising and recursion method based on residual error coding and decoding network |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN110223255A CN110223255A (en) | 2019-09-10 |
| CN110223255B true CN110223255B (en) | 2023-03-14 |
Family
ID=67816123
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910499262.XA Active CN110223255B (en) | 2019-06-11 | 2019-06-11 | Low-dose CT image denoising and recursion method based on residual error coding and decoding network |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110223255B (en) |
Families Citing this family (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110992295B (en) * | 2019-12-20 | 2022-04-19 | 电子科技大学 | Low-dose CT reconstruction method based on wavelet-RED convolution neural network |
| CN111325737B (en) * | 2020-02-28 | 2024-03-15 | 上海志唐健康科技有限公司 | Low-dose CT image processing method, device and computer equipment |
| CN111325695B (en) * | 2020-02-29 | 2023-04-07 | 深圳先进技术研究院 | Low-dose image enhancement method and system based on multi-dose grade and storage medium |
| CN111739114B (en) * | 2020-06-15 | 2023-12-15 | 大连理工大学 | Low-dose CT reconstruction method based on twin feedback network |
| WO2022226886A1 (en) * | 2021-04-29 | 2022-11-03 | 深圳高性能医疗器械国家研究院有限公司 | Image processing method based on transform domain denoising autoencoder as a priori |
| CN113256500B (en) * | 2021-07-02 | 2021-10-01 | 北京大学第三医院（北京大学第三临床医学院） | Deep learning neural network model system for multi-modal image synthesis |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107403419A (en) * | 2017-08-04 | 2017-11-28 | 深圳市唯特视科技有限公司 | A kind of low dose X-ray image de-noising method based on concatenated convolutional neutral net |
| CN108564553A (en) * | 2018-05-07 | 2018-09-21 | 南方医科大学 | Low-dose CT image noise suppression method based on convolutional neural networks |
Family Cites Families (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US10685429B2 (en) * | 2017-02-22 | 2020-06-16 | Siemens Healthcare Gmbh | Denoising medical images by learning sparse image representations with a deep unfolding approach |
| US11517197B2 (en) * | 2017-10-06 | 2022-12-06 | Canon Medical Systems Corporation | Apparatus and method for medical image reconstruction using deep learning for computed tomography (CT) image noise and artifacts reduction |
| KR20200063222A (en) * | 2017-10-09 | 2020-06-04 | 더 보드 어브 트러스티스 어브 더 리랜드 스탠포드 주니어 유니버시티 | Contrast dose reduction in medical imaging with deep learning |
| US10891762B2 (en) * | 2017-11-20 | 2021-01-12 | ClariPI Inc. | Apparatus and method for medical image denoising based on deep learning |
-
2019
- 2019-06-11 CN CN201910499262.XA patent/CN110223255B/en active Active
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107403419A (en) * | 2017-08-04 | 2017-11-28 | 深圳市唯特视科技有限公司 | A kind of low dose X-ray image de-noising method based on concatenated convolutional neutral net |
| CN108564553A (en) * | 2018-05-07 | 2018-09-21 | 南方医科大学 | Low-dose CT image noise suppression method based on convolutional neural networks |
Non-Patent Citations (3)
| Title |
|---|
| "Low-dose CT with a residual encoder-decoder convolutional neural network";H. Chen;《IEEE transactions on Medical Imaging》;20171231;全文 * |
| "低剂量CT的投影域去噪算法和后处理方法研究";崔学英;《中国博士学位论文全文数据库》;20150731;全文 * |
| "神经网络在CT重建方面应用的最新进展";方伟;《中国体视学与图像分析》;20190331;全文 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN110223255A (en) | 2019-09-10 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110223255B (en) | Low-dose CT image denoising and recursion method based on residual error coding and decoding network | |
| Kang et al. | Deep convolutional framelet denosing for low-dose CT via wavelet residual network | |
| CN110827216B (en) | Multi-generator generation countermeasure network learning method for image denoising | |
| Hu et al. | Artifact correction in low‐dose dental CT imaging using Wasserstein generative adversarial networks | |
| CN111325686B (en) | Low-dose PET three-dimensional reconstruction method based on deep learning | |
| CN108961237B (en) | Low-dose CT image decomposition method based on convolutional neural network | |
| Kuanar et al. | Low dose abdominal CT image reconstruction: An unsupervised learning based approach | |
| Tang et al. | Unpaired low-dose CT denoising network based on cycle-consistent generative adversarial network with prior image information | |
| CN111429379B (en) | Low-dose CT image denoising method and system based on self-supervision learning | |
| CN112258415B (en) | Chest X-ray film super-resolution and denoising method based on generation countermeasure network | |
| Huang et al. | Learning a deep CNN denoising approach using anatomical prior information implemented with attention mechanism for low-dose CT imaging on clinical patient data from multiple anatomical sites | |
| CN112396672B (en) | Sparse angle cone-beam CT image reconstruction method based on deep learning | |
| Zhang et al. | REDAEP: Robust and enhanced denoising autoencoding prior for sparse-view CT reconstruction | |
| Xue et al. | LCPR-Net: low-count PET image reconstruction using the domain transform and cycle-consistent generative adversarial networks | |
| Tao et al. | Learning to reconstruct CT images from the VVBP-tensor | |
| Amaranageswarao et al. | Wavelet based medical image super resolution using cross connected residual-in-dense grouped convolutional neural network | |
| Feng et al. | A preliminary study on projection denoising for low-dose CT imaging using modified dual-domain U-net | |
| CN112767273B (en) | Low-dose CT image restoration method and system applying feature decoupling | |
| CN112258438B (en) | LDCT image recovery method based on unpaired data | |
| Chan et al. | An attention-based deep convolutional neural network for ultra-sparse-view CT reconstruction | |
| Liu et al. | SDCNet: Smoothed dense-convolution network for restoring low-dose cerebral CT perfusion | |
| Park et al. | Low-dose CT image reconstruction with a deep learning prior | |
| CN111860836A (en) | Self-supervision learning method and application | |
| CN113436118B (en) | Low-dose CT image restoration method based on multi-scale convolutional coding network | |
| WO2023000244A1 (en) | Image processing method and system, and application of image processing method |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |